Research map / Architectures and efficiency / CNN design
Robust Learning with Progressive Data Expansion Against Spurious Correlation
Yihe Deng, Yu Yang, Baharan Mirzasoleiman, Quanquan Gu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add progressive data expansion stage after warm-up (ablation) | PDE · Waterbirds | worst-group accuracy: 86.0 -> 90.3 | Improved |
| add progressive data expansion stage after warm-up (ablation) | PDE · CelebA | worst-group accuracy: 87.8 -> 91.0 | Improved |
| add progressive data expansion stage after warm-up (ablation) | PDE · CivilComments | worst-group accuracy: 67.7 -> 71.5 | Improved |
| add all new data at once instead of gradually (ablation) | PDE · synthetic data | worst-group accuracy: 93.01 -> 74.24 | Hurt |
| reset momentum after warm-up (ablation) | PDE · synthetic data | worst-group accuracy: 93.01 -> 92.51 | Hurt |
| reset momentum after warm-up and add all data (ablation) | PDE · synthetic data | worst-group accuracy: 93.01 -> 67.69 | Hurt |
| expansion size 5 or 50 instead of 10 (ablation) | PDE · Waterbirds | worst-group accuracy: 90.3 -> 88.1 | Hurt |
| expansion learning rate 1e-2 instead of 1e-4 (ablation) | PDE · Waterbirds | worst-group accuracy: 90.3 -> 85.4 | Hurt |
| expansion learning rate 1e-3 or 1e-5 instead of 1e-4 (ablation) | PDE · Waterbirds | worst-group accuracy: 90.3 -> 89.4 | No clear effect |
| train GroupDRO with the larger learning rate used by PDE | GroupDRO · Waterbirds | worst-group accuracy: 86.7 -> 77.3 | Hurt |
| PDE vs GroupDRO convergence speed | PDE · Waterbirds | early-stopping epoch: 92 -> 8.9 | Improved |
| PDE vs GroupDRO | PDE · Waterbirds | average accuracy: 93.2 -> 92.4 | No clear effect |
Closest papers by meaning
- Let Samples Speak: Mitigating Spurious Correlation by Exploiting the Clusterness of Samples (CVPR 2025 · similarity 0.87)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.87)
- Severing Spurious Correlations with Data Pruning (ICLR 2025 · similarity 0.86)
- Feature Averaging: An Implicit Bias of Gradient Descent Leading to Non-Robustness in Neural Networks (ICLR 2025 · similarity 0.86)
- Label-Efficient Group Robustness via Out-of-Distribution Concept Curation (CVPR 2024 · similarity 0.85)
- The Group Robustness is in the Details: Revisiting Finetuning under Spurious Correlations (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
